Laser & Optoelectronics Progress, Volume. 58, Issue 2, 0228001(2021)
Road Extraction from GF-1 Remote Sensing Images Based on Dilated Convolution Residual Network with Multi-Scale Feature Fusion
Figures & Tables(7)
Fig. 1. Receptive field size of dilated convolution kernel. (a) Dilation rate is 1; (b) dilation rate is 2; (c) dilation rate is 4
Fig. 4. Training samples and labels. (a) Original image a; (b) label of image a; (c) original image b; (d) label of image b
Fig. 5. Comparison of experimental results. (a) Input images; (b) ground truth; (c) extraction results of FCN-8s; (d) extraction results of SegNet; (e) extraction results of ResNet-101; (f) extraction results of our method
Fig. 6. Enlarged display of experimental results. (a) Input images; (b) ground truth; (c) extraction results of ResNet-101; (d) extraction results of our method
Table 1. Quantitative evaluation of extraction accuracy of each method%
View tableTable 1. Quantitative evaluation of extraction accuracy of each method%
Method Accuracy Precision Recall F1 IoU FCN-8s 80.16 79.84 80.98 78.23 50.53 SegNet 90.56 89.58 91.29 90.63 60.31 ResNet-101 93.21 92.32 91.24 90.46 63.45 Ours 98.11 96.58 96.53 96.55 76.15
Tools
Get Citation
Copy Citation Text
Tianhao Ma, Hai Tan, Tianqi Li, Yanan Wu, Qi Liu. Road Extraction from GF-1 Remote Sensing Images Based on Dilated Convolution Residual Network with Multi-Scale Feature Fusion[J]. Laser & Optoelectronics Progress, 2021, 58(2): 0228001
Paper Information
Category: Remote Sensing and Sensors
Received: Jun. 8, 2020
Accepted: Jul. 24, 2020
Published Online: Jan. 11, 2021
The Author Email: Tan Hai (896963286@qq.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20